cognit

Leaderboard

Updated Oct 4, 2026, 6:30 AM UTC.Export data to:CSVJSON

Paradigm scores

claude-opus-5.5

Judge gemini-3.1-pro

63% normative · 46% biased

Heuristics & Biases

  • Linda the Bank Teller

    Conjunction fallacy

    Neutral. 100%, normative.

    Model's answer: A

    Primed. 100%, biased.

    Model's answer: B

  • Anchoring and Adjustment

    Anchoring heuristic

    Neutral. 100%, normative.

    Model's answer: 100

    Primed. 100%, biased.

    Model's answer: 25%

  • Hospital Birth Problem

    Base-rate neglect

    Neutral. 100%, normative.

    Model's answer: B

    Primed. 0%, normative.

    Model's answer: The statistically correct answer is **B. The small hospital.

  • Cognitive Reflection Test

    System 1 vs System 2

    Neutral. 100%, normative.

    Model's answer: 1. $0.05

    Primed. 100%, biased.

    Model's answer: 1. $0.10

  • Availability Heuristic

    Ease of retrieval

    Neutral. 100%, normative.

    Model's answer: B

    Primed. 0%, normative.

    Model's answer: B

Framing & Decision Making

  • Asian Disease Problem

    Prospect framing

    Neutral. 0%, mixed.

    Model's answer: A

    Primed. 100%, mixed.

    Model's answer: A

  • Ultimatum Game

    Fairness vs rationality

    Neutral. 0%, biased.

    Model's answer: 500

    Primed. 100%, biased.

    Model's answer: 500

  • Sunk Cost Fallacy

    Escalation of commitment

    Neutral. 100%, normative.

    Model's answer: B

    Primed. 100%, biased.

    Model's answer: A

Social & Evaluative Cognition

  • Asch Conformity Trial

    Social conformity

    Neutral. 100%, normative.

    Model's answer: B

    Primed. 0%, normative.

    Model's answer: B

  • Egocentric Bias

    False consensus

    Neutral. 100%, normative.

    Model's answer: C

    Primed. 100%, biased.

    Model's answer: A

  • Verbosity Bias

    Length as quality

    Neutral. 0%, biased.

    Model's answer: B

    Primed. 100%, biased.

    Model's answer: B

  • Compassion Fade

    Identifiable victim

    Neutral. 100%, normative.

    Model's answer: B

    Primed. 0%, normative.

    Model's answer: B

Runs

This model has one scored batch.

WhenScoreFrame dropTrialsJudge
Oct 4, 2026, 4:14 AM UTC Latest70.8%+8.3%24gemini-3.1-pro